47 lines
2.1 KiB
Python
47 lines
2.1 KiB
Python
|
import pandas as pd
|
|||
|
|
import numpy as np
|
|||
|
|
from matplotlib import pyplot as plt
|
|||
|
|
from sklearn import metrics
|
|||
|
|
from sklearn.model_selection import train_test_split
|
|||
|
|
from sklearn.neural_network import MLPRegressor
|
|||
|
|
|
|||
|
|
filein = "P:\\ULSTU\\ИИС\\Datasets\\heart_2020_norm.csv"
|
|||
|
|
|
|||
|
|
|
|||
|
|
# Метод обучения нейронной сети
|
|||
|
|
def reg_neural_net():
|
|||
|
|
df = pd.read_csv(filein, sep=',')
|
|||
|
|
x, y = [df.drop("HeartDisease", axis=1).values,
|
|||
|
|
df["HeartDisease"].values]
|
|||
|
|
|
|||
|
|
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.001, random_state=42)
|
|||
|
|
|
|||
|
|
mlp = MLPRegressor(hidden_layer_sizes=(100, 50), activation='tanh', solver='adam', random_state=15000)
|
|||
|
|
mlp.fit(x_train, y_train)
|
|||
|
|
y_predict = mlp.predict(x_test)
|
|||
|
|
err = pred_errors(y_predict, y_test)
|
|||
|
|
make_plots(y_test, y_predict, err[0], err[1], "Нейронная сеть")
|
|||
|
|
|
|||
|
|
|
|||
|
|
# Метод рассчёта ошибок
|
|||
|
|
def pred_errors(y_predict, y_test):
|
|||
|
|
mid_square = np.round(np.sqrt(metrics.mean_squared_error(y_test, y_predict)),3) # Рассчёт среднеквадратичной ошибки модели
|
|||
|
|
det_kp = np.round(metrics.r2_score(y_test, y_predict), 2) # Рассчёт коэфициента детерминации модели
|
|||
|
|
return mid_square, det_kp
|
|||
|
|
|
|||
|
|
|
|||
|
|
# Метод отрисовки графиков
|
|||
|
|
def make_plots(y_test, y_predict, mid_sqrt, det_kp, title):
|
|||
|
|
plt.plot(y_test, c="red", label="\"y\" исходная") # Создание графика исходной функции
|
|||
|
|
plt.plot(y_predict, c="green", label="\"y\" предсказанная \n"
|
|||
|
|
"Ср^2 = " + str(mid_sqrt) + "\n"
|
|||
|
|
"Кд = " + str(det_kp)) # Создание графика предсказанной функции
|
|||
|
|
plt.legend(loc='lower left')
|
|||
|
|
plt.title(title)
|
|||
|
|
plt.savefig('static/' + title + '.png')
|
|||
|
|
plt.close()
|
|||
|
|
|
|||
|
|
|
|||
|
|
if __name__ == '__main__':
|
|||
|
|
reg_neural_net()
|